Impact of fast-converging PEVD algorithms on broadband AoA estimation

Coutts, Fraser K. and Thompson, Keith and Weiss, Stephan and Proudler, Ian K. (2017) Impact of fast-converging PEVD algorithms on broadband AoA estimation. In: IEEE Sensor Signal Processing in Defence Conference, 2017-12-06 - 2017-12-07, London. (In Press)

[thumbnail of Coutts-etal-IEEE-SSPD-2017-Impact-of-fast-converging-PEVD-algorithms-on-broadband]
Text. Filename: Coutts_etal_IEEE_SSPD_2017_Impact_of_fast_converging_PEVD_algorithms_on_broadband.pdf
Accepted Author Manuscript

Download (951kB)| Preview


Polynomial matrix eigenvalue decomposition (PEVD) algorithms have been shown to enable a solution to the broadband angle of arrival (AoA) estimation problem. A parahermitian cross-spectral density (CSD) matrix can be generated from samples gathered by multiple array elements. The application of the PEVD to this CSD matrix leads to a paraunitary matrix which can be used within the spatio-spectral polynomial multiple signal classification (SSP-MUSIC) AoA estimation algorithm. Here, we demonstrate that the recent low-complexity divide-and-conquer sequential matrix diagonalisation (DC-SMD) algorithm, when paired with SSP-MUSIC, is able to provide superior AoA estimation versus traditional PEVD methods for the same algorithm execution time. We also provide results that quantify the performance trade-offs that DC-SMD offers for various algorithm parameters, and show that algorithm convergence speed can be increased at the expense of increased decomposition error and poorer AoA estimation performance.